Data governance
What is Data governance?
The discipline that governs how an organisation's data is owned, kept accurate and made safe to use: who is accountable, what quality it must meet, who is allowed to access it. For AI it is the foundation. Models are only as trustworthy as the data behind them, and most AI efforts stall on data long before they stall on models.
Why it matters
Data governance is not new, but AI has raised what is at stake in getting it wrong. Bad data used to produce a bad report someone could catch and correct. Now the same data trains a model, and its errors and gaps get baked into a system that repeats them at scale, on decisions that reach customers. The hard part is rarely technical. It is political. Governance means naming who is accountable for a dataset and holding them to a standard, and that steps on turf in organisations where data has drifted for years with no clear owner. Retrofitting it onto a live AI project is slow and painful, so the work belongs upstream, before the models arrive.
In practice
An AI project stalls when the team realises a critical dataset has no owner: its quality was never anyone’s job, and three departments each assumed another kept it clean. The modelling waits while that gets sorted, because a model built on it would inherit every inconsistency. Assigning ownership and a quality bar is what unblocks the project, and that step often decides whether it ships at all.
